Functions
template<typename Vector = DenseVector>
solver_return	turi::optimization::newton_method (second_order_opt_interface &model, const DenseVector &init_point, std::map< std::string, flexible_type > &opts, const std::shared_ptr< smooth_regularizer_interface > reg=NULL)

Detailed Description

Function Documentation

template<typename Vector = DenseVector>

solver_return turi::optimization::newton_method	(	second_order_opt_interface &	model,
		const DenseVector &	init_point,
		std::map< std::string, flexible_type > &	opts,
		const std::shared_ptr< smooth_regularizer_interface >	reg = `NULL`
	)

inline

Solve a second_order_optimization_interface model with a (dense) hessian Newton method.

Parameters

[in,out]	model	Model with second order optimization interface.
[in]	init_point	Starting point for the solver.
[in,out]	opts	Solver options.
[in]	reg	Shared ptr to an interface to a smooth regularizer.
[out]	stats	Solver return stats.

Template Parameters

Vector Sparse or dense gradient representation.

Note: The hessian is always computed as a dense matrix. Only gradients are allowed to be sparse. The implementation of Newton method must change when the hessian is sparse. I.e we can no longer perform an LDLT decomposition to invert the hessian matrix. We have to switch methods to Conjugate gradient or Sparse LDLT decomposition.

Definition at line 55 of file newton_method-inl.hpp.